Design Overview

Navigation hub for the Open Agent Harness design documents.

Terminology

• Terminology -- shared boundaries for Agent Engine, Agent Runtime, and Agent Spec
• Concept Relationships -- one map for Workspace, Worker, Sandbox, and Runtime

Three Core Concepts

Concept	Role	Description
Workspace	Capability boundary	Each workspace declares its own agents, models, tools, skills, actions, and hooks, and uses one consistent directory structure for discovery and execution.
Session	Context boundary	A continuous conversation or task collaboration, scoped to a workspace.
Run	Execution boundary	One model inference + tool loop. Runs are serial within a session.

Read by Topic

Architecture and Domain

• Architecture Overview -- layers, modules, request flow
• Domain Model -- core objects and relationships
• Storage Design -- PostgreSQL / Redis / SQLite responsibilities

Workspace Configuration

• Workspace Overview
• Settings | Agents | Models
• Skills | External Tools | Hooks

Engine

• Engine Overview
• Lifecycle | Context Engine
• Queue and Reliability | Events and Audit

External Interfaces

• API Reference | Schemas Overview

Deployment

• Quick Start | Deploy and Run | Server Config

Read by Role

Platform Engineers

1. Architecture Overview
2. Terminology
3. Concept Relationships
4. Domain Model
5. Workspace Overview
6. Engine Overview

Product / Integration Teams

1. Quick Start
2. Deploy and Run
3. API Reference
4. Streaming

Troubleshooting

1. Deploy and Run
2. Lifecycle
3. Queue and Reliability
4. Events and Audit

Translation Note

Not every page has an English translation yet. When no English page exists, the site falls back to the Chinese source.